This model works better with a 1.3 strengh, I also recommend using the "anime" tag for best results and to reinforce the style,
lora.TA_trained (6).safetensors